Journal of Bioteehnology, 10 (1989) 277-284
277
Elsevier JBT 00387
Concentration of L-phenylalanine with a reverse osmosis m e m b r a n e W.C. M c G r e g o r XOMA Corporation, 2910 7th Street, Berkeley, CA 94710, U.S.A.
(Received14 February 1989; accepted 30 March 1989)
Summary A high flux, thin film composite reverse osmosis (RO) membrane was used to concentrate L-phenylalanine (L-Phe) from clarified bioreactor harvest media. At p H 10 + 0.5 and 50°C, concentrations of 100 g 1-1 were easily achieved and at fluxes from 17 to 119 1 m -2 h -1. Rejection coefficient for L-Phe was inversely proportional (as the log) to retentate concentration. A preliminary system study showed that stages in a cascade could be used to recover essentially all of the product from clarified harvests. The study shows the importance of empirical evaluation as the basis of design and suggests that bioprocess applications of RO are likely to be case specific. Phenylalanine, L-; Reverse osmosis; Membrane; Amino acid; Flux decfine; System design; Membrane separation; Isolation
Introduction The amino acid, L-phenylalanine, is economically important because of its primary role in aspartame (Nutrasweet TM) synthesis (Klausner, 1985). One of the methods for manufacture of L-phenylalanine is the conversion of t r a n s - c i n n a m i c acid by the enzyme, phenylalanine ammonia lyase. Conversion can take place at high pH in a bioreactor containing the substrate and enzyme-rich cells of the yeast Rhodotorula rubra (Hamilton et al., 1985). Correspondence to: W. McGregor,XOMA Corp., 2910 7th St., Berkeley,CA 94710, U.S.A.
Work performed in the Cell Products (Squibb) pilot plant in New Brunswick, NJ (1984). 0168-1656/89/$03.50 © 1989 ElsevierScience Publishers B.V. (BiomedicalDivision)
278 The product in the harvest stream from the bioreactor may be relatively high in concentration (as much as 59 g 1-1 have been reported (by Hamilton et al., 1985) and recovered by simple crystallization. In some cases the overall recovery can be greatly increased by first concentrating the product at a high temperature and high p H before crystallization. This paper describes preliminary work on the concentration of L-phenylalanine (L-Phe) in clarified bioreactor harvest streams using reverse osmosis (RO) as an alternative to evaporation.
Methods
Reverse osmosis experiments were performed with samples from a proprietary, commercial industrial process briefly described as follows: industrial scale bioreactors ( > 50 000 1) were operated with a strain of Rhodotorula rubra suspended in a solution of cinnamic acid and ammonia at p H 10.3. After conversion, cells were removed from the harvest in a nozzle centrifuge (Westfalia) and the supernatant was clarified by ultrafiltration (UF) with Romicon hollow fibers (100000 MWCO). Samples of the permeate were used for pilot RO experiments either directly or after ammonia (but not water) was partially removed by evaporation (on industrial scale). Concentration experiments were conducted on a DDS Lab 20 Unit which was operated with three RO40 thin film composite reverse osmosis membranes (Bio-Recovery Inc., Northvale, N J) in the stack. Each membrane was 0.036 m 2. Retentate was recirculated at approximately 11.3 1 rain-1 with a high pressure piston pump. Feedstreams (typically 20 1) were initially at 50 ° C and were generally maintained at 5 0 ° C through pumping energy but dropped slightly because experiments were performed at room temperature. Flux data were corrected for any temperature drops. Assays for L-phenylalanine were by H P L C on a 0.45 cm × 30 cm Supelco column with C18 reverse phase, 5 ~tm particles. The mobile phase was 0.1 M potassium phosphate in methanol (9 parts buffer to 1 part methanol). The p H of diluted samples for HPLC was adjusted with ammonium hydroxide or sulfuric acid.
Results
Solubility of L-phenylalanine in water was determined as a function of p H and temperature (Table 1). This information was useful in determining the limits and best conditions for RO operation. The objective of subsequent RO experiments was to concentrate L-phenylalanine at high pH, approximately p H 10, and at temperatures of approximately 50 ° C. Membranes were characterized with tap water and with permeate from the Romicon hollow fiber membrane system. Results of flux at various temperatures for three different membranes are shown in Fig. la,b. Slight membrane-to-membrane differences were noted so averages from these plots were used to correct subsequent feed stream flux data to 50 ° C and to check for restoration of water flux. Mem-
279 TABLE 1 SOLUBILITY(g 1-1) OF L-PHENYLALANINEIN WATER * pH
Temperature(°C) 70
60
50
25
4
1.5 2.0 5.5 7.0 7.5 9.5
128 87 50 62 79 324
128 87 44 50 63 206
74 56 39 40 49 149
62 52 30 32 40 132
56 24 25 26 46
* pH of samples adjusted with ammoniumhydroxide or sulfuric acid. Published solubility values (Merck Index, 10th edn, 1983) in water without specified pH are: 19.8 g 1-1 at 0 ° C, 29.6 g 1-1 at 25 ° C, 44.3 g 1-1 at 50 o C, 66.2 g 1-1 at 75 o C, and 99.0 g 1-1 at 100 o C. branes were cleaned between runs to restore water flux by circulating a 0.5% solution of P3 Ultrasil 10 (Henkel Corp., Chemical Division), at 4 5 ° C for approximately 30 rain. A flush-clean-rinse cycle was repeated if necessary to restore water flux. Fig. 2 shows the effect of retentate concentration of L-Phe on the rejection coefficient. Since it was clear that the rejection coefficient was inversely proportional (as the log) to the retentate concentration, a system design could probably be developed with stages. Permeate from the first stage could be fed directly to the second stage and so on, and by such a cascade, essentially all of the product could be recovered (Fig. 3). Flux improved by removing part of the ammonia. Fig. 4 shows flux decay curves for cascaded stages with and without ammonia partially removed.
Discussion
This report documents an example of amino acid concentration from a clarified bioreactor harvest by reverse osmosis. Although the concept of small molecular weight bioproduct concentration by reverse osmosis has been obvious for many years, published experience is scarce (Dwyer, 1987). Experience from this study reinforces the importance of comparative testing in membrane selection. The DDS unit was a convenient system for this purpose; several different membranes could be compared at once. Essentially the same factors that are important in ultrafiltration membrane selection (McGregor, 1986) were also important in this study with reverse osmosis membranes: membrane manufacturer, molecular rejection rating, membrane composition and flux. In the present study, a membrane composition that would tolerate high ammonia at p H 10 was foremost in importance. In addition, a relatively high salt rejection rating was sought because of the low molecular weight of L-Phe (165). High flux is always desirable. The comparison of membranes with different rejection ratings and
280 255
o []
238
o
221 ~ '~
204 187 170 153 136192 2O
I
I
25
I
1
30
35
1
40
TEMPERATURE
45
50
(*C)
51.0 b
[3
476442408"
[]
374-
,~
[]
[]
34 0" 'E
306272. 238,
204, 170. 13.6 20
, 22
24
26
28
30
32
34
36
38
40 42
44
46
48
50
TEMPERATURE (°C)
Fig. l. (a) Effect of temperature on RO flux of tap water through RO40 membranes 1 m- 2 h- 1. Pressure at 375:2 bar. m, plate 1; zx, plate 2; o, plate 3; average flux decay with temperature drop was approximately 3.4 1 m -2 h-1 per o C. (b) Effect of temperature on RO flux of L-phenylalanine in clarified bioreaetor harvest fluid. Feedstream was permeate from PM-100 UF membranes. Pressure at 37 + 2 bar. n, plate 1; zx, plate 2; o, plate 3; average flux decay with temperature drop was approximately 0.85 1 m -2 h -1 per °C.
from different manufacturers, s h o w e d as m u c h as a 10-fold difference in flux (data n o t shown). Even t h o u g h the h i g h flux o f the R O 4 0 m e m b r a n e s was s o m e w h a t offset b y the l o w rejection (salt rejection rating of 40%), a cascaded s y s t e m of stages was readily apparent. Preliminary design was empirically derived since rejection coefficients for different stages could not be predicted from basic performance data. Fig. 3 shows that the rejection coefficient at a given L-Phe concentration is different in each stage, probably because of differences in f o u l i n g or secondary m e m b r a n e f o r m a t i o n
281 1oO0 0
LU U LL LL W o ij
0
E) b w w o~
0.1
I 20
0
4
L-Phe
I0
I 60
CONCENTRATION
IN
I 80
1 100
RETENTATE
120
( g l -~ )
Fig. 2. Effect of retentate concentration (L-Phe) on rejection coefficient. Rejection coefficient = (C R Cp)//CR, where C R = L-Phe concentration in the retentate; Cp = L-Phe concentration in the permeate. Feedstream containing a m m o n i a at p H 10.5. Pressure at 37 + 2 bar.
by other components. Even so, at low L-Phe concentrations in the relatively clean 3rd stage, essentially all of the product was retained. In this study, the same membrane stack, after cleaning to restore water flux, was used for each stage. Operation of the 3rd stage was relatively short because only about 7 1 of feed were available. System hold-up did not allow for more than the initial flux and rejection data as shown in Figs. 3 and 4. Fig. 4 shows that higher fluxes could be achieved after partial removal of ammonia by evaporation. In an actual design, the advantage
q:P
lO 09
~w 0
08
b_ b_ U
~) 0 6 m
0.5 0
I 10
I 20 L-Phe
I 30
I 40
I 50
CONCENTRATION
I 60
I 70
IN R E T E N T A T E
I 80
I 90
100
(g1-1)
Fig. 3. Effect of retentate concentration (L-Phe) on rejection coefficient in stages of a cascade, o , stage 1; zx, stage 2; O, stage 3, a m m o n i a partially removed from feedstream, p H 9.5. Pressure at 37 + 2 bar. Flux data and initial concentrations shown in Fig. 4.
282 170" 153" 136" 119"
V
V__
V
- - 0 - 0--O--D 0
~2'
-
4)
, 0 ~
102. 85"
o/ E
68" 51 34
LL 17. 0
I
I
I
I
I
I
10
15
20
25
30
35
aO
CONCENTRATION FACTOR, BY VOLUME
Fig. 4. Flux decay during reverse osmosis concentration of L-Phe at different stages of a cascade system. ©, stage 1 with ammonia at pH 10.5, [L-Phe]0= 54 g 1-1; zx, stage 2 with ammonia at pH 10.5, [L-Phe]0= 38 g 1-1; E], stage 1 with ammonia partially removed at pH 9.5, [L-Phe]0= 43 g 1-1; ~, stage 2 with ammonia partially removed at pH 9.5, [L-Phe]0=16 g 1-1; v, stage 3 with ammonia partially removed at [L-Phe]0 = 5 g 1-1. Pressure at 37 + 2 bar. of higher flux at lower ammonia concentrations would have to be weighed against the cost of evaporation. It is important to emphasize that the system study was preliminary and did not result in an actual design. The Cell Products Corporation was closed down before additional data could be collected for final design and for proper economic evaluation. Even though the project was not completed, the work is presented for whatever modest value it may have. The molecular complexity of most industrial bioprocess streams (of which this is an example) together with the wide variety of available but inadequately characterized membranes means that membrane separations of this type will be mostly empirically derived and case specific for at least the near future.
Acknowledgements Assistance with assay data by Michael Sakelarides is gratefully acknowledged. I thank Larry Niesen and Paul Hellman of Bio-Recovery Inc., for membrane samples and for assistance in preliminary design. The type of support provided by Bio-Recovery greatly enhances empirical studies such as described here. Margot Stanton, Kily Hailer and John Thrift were most helpful in preparing the manuscript.
283
References Dwyer, J.L. (1987) Small molecule membrane separation processes. Biopharm (Sept.) 71-75. Hamilton, B.K., Hsiao, H.-Y., Swann, W.E., Anderson, D.M. and Delente, J.J. (1985) Manufacture of L-amino acids with bioreactors. Trends Biotechnol. 3, 64-68. Klausner, A. (1985) Building for success in phenylalanine. Biotechnology 3, 301-307. McGregor, W.C. (1986) Selection and use of ultrafiltration membranes. In: McGregor, W.C. (Ed.), Membrane Separations in Biotechnology, Marcel Dekker, New York, pp. 1-36.